Xiaocong Du

About

Xiaocong Du is a researcher focused on advancing continual learning and trajectory prediction for dynamic, real-world systems like autonomous vehicles, surveillance drones, and robotics. Their most impactful contribution is the development of **Progressive Segmented Training (PST)** for single-network continual learning, introduced in their highly cited 2019 paper (13 citations). PST addresses the critical challenge of enabling a single neural network to learn from a data stream without catastrophic forgetting, preserving prior knowledge while adapting to new tasks—a fundamental requirement for lifelong learning in embodied AI. Du also pioneered **DAT-RNN**, a diverse attention mechanism for trajectory prediction (2020, 2 citations), which improves the modeling of complex, multi-agent spatial-temporal interactions. By tackling both the retention of learned skills and the accurate forecasting of future paths, Du’s work directly supports the deployment of safer, more adaptive autonomous systems. Their research bridges core machine learning challenges with pressing engineering needs, making them a notable contributor to the fields of continual learning and intelligent dynamic systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Single-Net Continual Learning with Progressive Segmented Training

   13 citations · 2019
2. 2
   Single-Net Continual Learning with Progressive Segmented Training (PST)

   4 citations · 2019
3. 3
   DAT-RNN: Trajectory Prediction with Diverse Attention

   2 citations · 2020

Key Collaborators

Contact & Links